---

FilterGS utilizes a single RTX 4090 for training highly realistic urban-scale models and for their real-time rendering. Visit our project page for more demos.

Our code is built upon PyTorch and leverages gaussian-splatting and LoG techniques.

Quick Start & Dataset Preparation

For a smooth setup, follow the installation guide.

We employ Colmap to prepare the dataset. Refer to the preprocessing documentation for detailed instructions.

Training

Training the model is as simple as one command:

python3 apps/train.py --cfg config/GauUScene/college/train.yml split train

We automatically configure heuristic parameters based on the dataset size.

Rendering

Before rendering the model, it is necessary to calculate the ancestor_path attribute of the model, which will take about 5 minutes:

python apps/precompute_ancestor.py --ckpt /PATH/TO/YOUR/MODEL.pth --out /OUTPUT/PATH/model_ancestor.pth

Rendering the model is also as simple as one command:

python3 apps/render.py --cfg config/GauUScene/college/render.yml split train

--skip-save and --debug are optional parameters; --skip-save skips saving rendered images, while --debug outputs detailed parameters of the rendering process.

Acknowledgements

We acknowledge the following inspirational prior work:

• gaussian-splatting
• Level of Gaussian
• MatrixCity dataset
• UrbanScene dataset
• GauUScene dataset

More details to be added later. If you have any questions, please feel free to point them out in the issue section.

Contributions are warmly welcomed! If you've made significant progress on any of these fronts, please consider submitting a pull request.

Citation

Our paper will be available within a week. We greatly appreciate your early interest in this work!